Developing Frameworks and Modeling Behavior for Cognitive Sonar
Abstract
Funding is provided to develop and assess processing architectures and models for incorporating high-level decision making, or cognitive processing, for sensing in active sonar systems. The focus of the proposed effort is the exploration of models for incorporating high-level decision-making, or cognitive processing, in active sonar systems. The aim is to leverage existing work in the artificial intelligence, optimization, and radar realms to develop potential models applicable to high-level sonar automation. While existing active sonar platforms may allow operators to control specific parameters such as ping rate and waveform, they do not support adaptation to higher-level requests from operators, e.g. a request to more closely monitor a particular target in track or a specified region of the surveillance space. We propose to develop frameworks for systems that translate highlevel tasks to a set of parameter adaptations and arbitrate among competing tasks and demands. Development of a cognitive sonar system will advance the Navy’s ability to reliably conduct surveillance and tracking by allowing sonar operators to interact with the system via high- level commands, thereby reducing the time required to identify a threat. We consider existing research in both cognitive systems and artificial intelligence to identify frameworks suitable for application to the cognitive sonar problem. We propose to explore three possible models for cognitive sonar: stochastic control techniques applied to a partially observable Markov decision process, a Bayesian-filtering based optimization model, and a goal-directed autonomy model for goal reasoning. For each possible model, we propose to specialize the existing framework to cognitive sonar and to evaluate the performance and complexity of the result, in particular assessing whether the computational requirements are practical for complex surveillance problems. The proposed work will be performed in collaboration with researchers at the Naval Undersea Warfare Center (NUWC). The PI will spend Spring 2017 working closely with NUWC as the focus of a sabbatical leave. The collaboration with NUWC will support the development of improved cognitive sonar models with direct application to Navy problems, as well as assessment of the proposed models on real data sets and highfidelity simulated data sets. A main goal of this collaborative effort is to develop the foundation for ongoing collaborations between researchers at NUWC and at George Mason University. To that end, NUWC will pursue a CRADA with George Mason University to facilitate long-term collaboration.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Nov 23, 2016
- Source ID
- N000141612494
Entities
People
- Jill Nelson
Organizations
- George Mason University
- Office of Naval Research
- United States Navy